Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

[Patman]

Yes, but doesn't that apply just for 2-way ties? I don't believe that HEA uses a round-robin of H2H results (as I think the ECAC still does) to break 3- and 4- etc., way ties, but sorts first by total league wins to winnow the field to group(s) of two, and then breaks that tie by the H2H record between just those two teams. That's how I remember it anyways, 'tho I could be totally out to lunch ...

That's ok, hockey east has never been clear on the exact way of breaking ties... If a worst case of sequencing issues came up they would be stuck and it would probably not end well.

Personally, due to clarity, I've always preferred the NFL style where you determine the teams in order and once a tie is broken amongst the several if there are still ties at the top you start the process over.

May not produce the best result, but at least you know the result.

[RENCEB]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

So does this mean that MC @ BU is, in effect, a play-in game? Both are buried deep in the PWR and, it would seem that a loss by either team basically requires a HE tourney title for a trip to the NCAA tourney. I'm guessing the math does not work otherwise.

I'd bet that both teams are already at the point where the HE title is the only way they get in. MC's loss to BC was devastating from a PWR perspective; would have been huge to pick up H2H comparision points vs. top 5 team. BU getting swept by UML may have been a nail in their PWR coffin.

[Umileated]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

Jim Dahl does some awesome break downs over on his Sioux Sports Talk blog. If you scroll down, you'll find projections for BU and Providence. I don't believe he takes conference tournaments into account, but I could be wrong.

http://blog.siouxsports.com/2013/02/...rnament-berth/

This was of course how BU looked going into being swept by Lowell.

[Todd]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

Yes, but doesn't that apply just for 2-way ties? I don't believe that HEA uses a round-robin of H2H results (as I think the ECAC still does) to break 3- and 4- etc., way ties, but sorts first by total league wins to winnow the field to group(s) of two, and then breaks that tie by the H2H record between just those two teams. That's how I remember it anyways, 'tho I could be totally out to lunch ...

No. That's how they begin to settle all ties, regardless of how many are tied.

That's ok, hockey east has never been clear on the exact way of breaking ties... If a worst case of sequencing issues came up they would be stuck and it would probably not end well.

Personally, due to clarity, I've always preferred the NFL style where you determine the teams in order and once a tie is broken amongst the several if there are still ties at the top you start the process over.

May not produce the best result, but at least you know the result.

Actually, that's what they do.

It's mostly clear in their footer of the press releases and on their website - and, of course, in every detailed breakout of every tie I resolve in posts on this thread for the last at least... seven years.

The only ambiguity through the years - as has been discussed many times here over several seasons and at least once this season - has been how to apply the H2H records to sort out who to keep or toss and in what order. They had been consistent one way, and then (inadvertently) switched to another way when predicting in the only year where it mattered - the only more-than-two way tie (so far) was for 3rd in 2010. They have committed to the new way.

The only imprecision was in their description of the process for multi-way H2H. They have always used H2H as the first tb, at least as far back as I can remember their procedures.

[Todd]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

Playing with the what-if simulator for Hockey East, there is a possibility, however very remote, that the top 6 teams finish with 32 points each.


It would require the following results with points taken/available noted.

BC:
Lose to Lowell 0/2
Split with Providence 2/4
Split with Vermont 2/4

BU:
Lose to Merrimack 0/2
Sweep Vermont 4/4
Sweep Northeastern 4/4

MC:
Beat BU 2/2
Split with Lowell 2/4
Take only one point from Amherst 1/4

UML:
Beat BC 2/2
Split with Merrimack 2/4
Split with Providence 2/4

PC:
Split with BC 2/4
Split with Lowell 2/4

UNH:
Split with Amherst 2/4
Split with Maine 2/4

That's awesome.

There's a little wiggle room in there, too, among this Group of Six (G6).

For example:
Since UNH's games are outside the group, they just need four points, regardless of how they come from UMA and ME. Sweep/Swept, split/split, ... however. Just so long as it's four total.

Similarly, BC/PC/UML all play each other, so the points there are a little fluid. Within these three, you have BC losing to UML and splitting w/ PC, PC taking two from each, and UML also getting two from each. So, if, say, BC beats UML, that doesn't mess the whole thing up. Instead UML can get their four points by sweeping PC. PC then gets their four points by sweeping BC. BC already has their two points from this trio, so they don't need the two from PC. Status quo maintained. Only the tbs are affected.

If any ties get thrown in there, it just means we need to account for a round robin of ties to even everything back up - both in point distribution and even-odd totals.

You can also factor in that:
BC has UVM for spillover (in the way that UNH can get their points form wherever because it doesn't impact the G6).
UML has spillover to MC, who can spill to UMA.
UML, via MC, can also spill to BU who has points to play with with UVM and NU.

What I mean by the spillover factors is that if...
MC and PC both shut down BC, the Eagles could still get four points from UVM to hit 32. The UVM points don't impact the G6.
That would leave PC with two too many points, and the BC games are accounted for, so they would need to lose out to UML.
Now UML has two too many, with BC and PC done, so they must lose to MC to stay at 32.
MC now has two extra. They can't drop two to UMA, because they're only taking one. They could drop one more to UMA (getting swept) and one to BU, or both extras to BU.
Whatever extra points MC shoves over to UMA are out of the G6.
Whatever they shunt over to BU are points that BU doesn't take from either (or both, if two points) of UVM/NU, which then drop out of the G6.

Everyone in the G6 still ends up at 32.

So... wiggle room.

[Felger]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

As unpredictable as this great race is, imagine how it would be if we had the screwball NHL system, where some games are worth 2 points and some are worth 3.

Then again, if we had that convoluted system, the race might not be this way....

[WrathOfAramark]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

wiggle room

That's even more awesome.

I hope Joe Bertagna reads this and his head explodes.

[Go-UML]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

i hope joe bertagna reads this and his head explodes.

:d

[Todd]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

After Sun 2/24:
BC 2 @ MC 1 OT

--- Home Lock - 35 (PC/UNH/MC/BC) ---
PC 28 - 36 [1-6]
BC 28 - 38 [1-6]
UNH 28 - 36 [1-6]
MC 27 - 37 [1-6]
UML 26 - 36 [1-7]
BU 24 - 34 [1-8]
--- Home Eligible - 28 (Top 5) ---
--- In - 24 (ME) ---
UVM 19 - 27 [5-10]
UMA 17 - 25 [6-10]
ME 16 - 24 [7-10]
NU 13 - 21 [7-10]
--- Out - 17 (UMA) ---

Remaining LEAGUE storm-adjusted schedules:
PC - @BC/BC, @UML/UML
BC - UML, PC/@PC, @UVMx2
UNH - UMAx2, MEx2
MC - @BU, @UML/UML, UMA/@UMA
UML - @BC, MC/@MC, PC/@PC
BU - MC, UVMx2, @NU/NU
UVM - @BUx2, BCx2
UMA - @UNHx2, @MC/MC
ME - NUx2, @UNHx2
NU - @MEx2, BU/@BU


Well, either MC or BC had to be in first after Sunday's game was over.

As it happened, it took OT to decide that it would be BC jumping up from 4th to 1st, leaving MC a point behind the leaders. BC can also now be no lower than 6th.

----
Any bets on UML beating BC and MC tying BU Tuesday night to give us five-way split at 28 with two weekends to go?

----
After Tuesday, all the games-in-hand and storm adjustments will be done and everyone will have four games left.

Four?

How can this much be left to decide with only four games left? I know Tuesday will change something, but still... four?

As it turns out, the year where the top four were all facing off on the final weekend and any of them had a shot at the top seed (noted earlier in this thread) was in the final year of the nine-team format (04-05). Still, while we didn't know who would specifically be first through fourth, we knew that those would be the four with Home Ice.

Here we are in the final year of the ten-team format and I'm reasonably sure that it has never happened that no one has clinched Home Ice with this little time left. Hell, with a five-(or six-, per WrathOfAramark)way tie, you'd have a team (or teams!) that can hang a banner as league (co-)champs that wouldn't get Home Ice.

Should I point out that next year will be the only, and hence final, year of the eleven-team format?

----
Since it was just the one game between teams at the top, the In and Out lines aren't affected.

----
Also, since it was just the one game between teams in the Home Lock indicator, the way it is bunched this year, that line also does not move.

----
However, since BC jumped up from 26 to 28, that makes MC (27) and UML (26) the current 4/5. With four points still to split between the two, that means that at least one of them will have to reach 29 points.

Since UNH could sit at 28 with a schedule outside of the top teams, the MC/UML loser (or UML in a 2/2 split) could stay at 28 or below, and BU has shown that they can stay right where they are, that sets the Home Eligible line at 28.

----
For i-dotting and t-crossing:
If they were seeded today, with a pair of BC/PC games pending, the RR at the top should fall PC (2-0-2), BC (1-1-2), UNH (1-3-2). In case they get funky and go back to the old way of "bottom up" breaks, BC and PC are currently split H2H (0-0-1) and BC has more wins than PC, so BC would top out. But that's not what they've told us will happen, so I'll stick with PC/BC/UNH.

[Todd]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

That's even more awesome.

I hope Joe Bertagna reads this and his head explodes.

I wish he had been reading this.

I'm sure, given where we are, that this is going to come up even more repeatedly as we draw closer. There are a couple of notes made in various posts earlier in this thread, but for those that really want to follow the discussion, the link above is from the end of 09-10, when all of this really mattered. I don't think I included all of my emails with the league office in there, but ... well rather than revisit the whole thing here, that's what the link is for.

If anyone has Qs, feel free to ask and any of us that were in that discussion that are still reading here can assist.

----
Edit:

Actually, I did include the emails.

Here is the link to Post 100, which summarizes the issue, past and present.

When I later spoke to Joe in person, he indicated that they were doing things the new way (he didn't say "new", that's what we called it in our discussions) going forward. I think he was aware of the situation that had come up, but I'm not positive after a couple of years have gone by.

I don't recall if I ever did hear back from Pete, although I think I tried him again at the beginning of the next season. Not sure if he ever understood the difference.

jjmc85's post 112 is similar to something that I had noticed about BU having "clinched" when they shouldn't have under the rules that had existed up to that point. Caused quite a bit of discussion, as you can see.

[Todd]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

So does this mean that MC @ BU is, in effect, a play-in game? Both are buried deep in the PWR and, it would seem that a loss by either team basically requires a HE tourney title for a trip to the NCAA tourney. I'm guessing the math does not work otherwise.

I'd bet that both teams are already at the point where the HE title is the only way they get in. MC's loss to BC was devastating from a PWR perspective; would have been huge to pick up H2H comparision points vs. top 5 team. BU getting swept by UML may have been a nail in their PWR coffin.

Jim Dahl does some awesome break downs over on his Sioux Sports Talk blog. If you scroll down, you'll find projections for BU and Providence. I don't believe he takes conference tournaments into account, but I could be wrong.

http://blog.siouxsports.com/2013/02/...rnament-berth/

This was of course how BU looked going into being swept by Lowell.

That's a great use of data viz for sports.

Jim Dahl's analysis goes through the end of the RS. As he shows, at that point, if BU won 5 of its remaining 7 (it has already lost 2 to UML, so has only 5 left), they'd average out around a 13 PWR going into HE playoffs. Win "only" four, and they're averaging right on/under the bubble around 16-17.

What I don't know is whether he was comparing all the COP, TUC, H2H issues, or just RPI.

In the case of HE 3-6, they will also be playing strong enough competition that it will change their TUC record. Whoever survives will probably get that again in the semis (and perhaps the final). Win or lose, it will increase the strength of schedule for all in that group as well.

The other thing to note is that TUC records near the bubble - really everywhere but perhaps the very top teams - tend towards .500.

MC lost to BC and so lost a number of pairs. If they then beat BU or UML or a TUC in the HEQFs and beyond, those pairs can come right back.

Also to be considered: at this point in the season, RPIs tend to fall faster (or rise higher) towards .5000 than they climb (or fall) away from it. The further from the middle you are, the bigger the jump towards it and the harder the pull away.

Of course, the other teams in the other leagues also have to win and lose games, and teams will join and drop off of the TUC list as well. It's possible that MC's loss to Alaska or tie with Colgate could drop off their TUC tally. If UVM sweeps BU and BC, they should top .5000, adding a 2-0-1 to MC's TUC - and so on.

In sum: while we're drawing close to the time when it matters, there's more under the surface than just the games a given team is playing, so there's volatility - esp. at the bubble.

Umileated reminded me about Dahl's RPI page (my browser had dumped my links a while back). If you're into the math, from there, you can drill down into the detail of how any remaining games impact any teams RPI.

Example: The most important game for BU to win, in terms of gaining RPI, is the MC game. The most important not to lose, in terms of damaging their RPI, are the remaining UVM/NU games. If they sweep out the RS, they'll pick up about .0140 points. If they get 3/5, they'll be about where they are.

MC could add about .0160 by sweeping, and could be a little ahead of where they are now with 3/5.

Anyway, neither are "must" wins, but there are so few games left that if you don't win soon, you have no way to change where you are.

Insert Bill Parcells, Dennis Green, or - of course - Jim Mora quote of your choosing here.

[Todd]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

As unpredictable as this great race is, imagine how it would be if we had the screwball NHL system, where some games are worth 2 points and some are worth 3.

Then again, if we had that convoluted system, the race might not be this way....

Here and here - aka the Shootout Years.

[jeteye1717]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

For i-dotting and t-crossing:
If they were seeded today, with a pair of BC/PC games pending, the RR at the top should fall PC (2-0-2), BC (1-1-2), UNH (1-3-2). In case they get funky and go back to the old way of "bottom up" breaks, BC and PC are currently split H2H (0-0-1) and BC has more wins than PC, so BC would top out. But that's not what they've told us will happen, so I'll stick with PC/BC/UNH.

OK, help me out here, as it appears I mis-remembered how HEA breaks standings ties of 3+ teams or more ... Since Todd has convinced me that cumulative H2H records amongst the tied teams is the first criterion (not most wins, as I mistakenly assumed), I can see how the above RR ordering is calculated. But once PC is deemed to top out, shouldn't the H2H of records of *just* the two remaining teams (BC, UNH) be used to break the tie for spots 2 and 3, without including PC who's already been awarded the top seed? And since that record is tied at 1-1-1, wouldn't most wins between just those two teams kick in next? If PC is deemed #1, then they're no longer tied with teams below them; thus only H2H records involving the remaining tied teams should be used. Or so I'd assumed from a strictly logical parsing of the situation, which of course means it won't and doesn't happen that way :-( Why does my brain keep flashing back - and not in a good way - to the infamous 'infinite loop' that once bedeviled the ECAC tie-breaking system ...

[Umileated]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

OK, help me out here, as it appears I mis-remembered how HEA breaks standings ties of 3+ teams or more ... Since Todd has convinced me that cumulative H2H records amongst the tied teams is the first criterion (not most wins, as I mistakenly assumed), I can see how the above RR ordering is calculated. But once PC is deemed to top out, shouldn't the H2H of records of *just* the two remaining teams (BC, UNH) be used to break the tie for spots 2 and 3, without including PC who's already been awarded the top seed? And since that record is tied at 1-1-1, wouldn't most wins between just those two teams kick in next? If PC is deemed #1, then they're no longer tied with teams below them; thus only H2H records involving the remaining tied teams should be used. Or so I'd assumed from a strictly logical parsing of the situation, which of course means it won't and doesn't happen that way :-( Why does my brain keep flashing back - and not in a good way - to the infamous 'infinite loop' that once bedeviled the ECAC tie-breaking system ...

It appears that Lowell has just joined the party, just to complicate things further.

[UNH09]

Re: Hockey East - Who's in, who's out, who's home: by the numbers - 2012-13 edition

It appears that Lowell has just joined the party, just to complicate things further.

Lowell is out, given the record against BC and UNH.